Planning and Optimization
A3. Transition Systems and Propositional Logic
Gabriele R¨ oger and Thomas Keller
Universit¨ at Basel
October 1, 2018
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 1 / 27
Planning and Optimization
October 1, 2018 — A3. Transition Systems and Propositional Logic
A3.1 Transition Systems A3.2 Example: Blocks World
A3.3 Reminder: Propositional Logic A3.4 Summary
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 2 / 27
Content of this Course
Planning
Classical
Tasks Progression/
Regression Complexity Heuristics
Probabilistic
MDPs Uninformed Search
Heuristic Search Monte-Carlo
Methods
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 3 / 27
Goals for Today
Today:
I introduce a mathematical model for planning tasks:
transition systems Chapter A3
I introduce compact representations for planning tasks suitable as input for planning algorithms
Chapter A4
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 4 / 27
A3. Transition Systems and Propositional Logic Transition Systems
A3.1 Transition Systems
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 5 / 27
A3. Transition Systems and Propositional Logic Transition Systems
Transition System Example
Transition systems are often depicted as directed arc-labeled graphs with decorations to indicate the initial state and goal states.
`
1`
1`
1`
1`
3`
3`
2`
4`
3`
4`
4`
4`
2`
2c(`
1) = 1, c(`
2) = 1, c (`
3) = 5, c(`
4) = 0
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 6 / 27
A3. Transition Systems and Propositional Logic Transition Systems
Transition Systems
Definition (Transition System)
A transition system is a 6-tuple T = hS , L, c, T , s 0 , S ? i where
I S is a finite set of states,
I L is a finite set of (transition) labels,
I c : L → R + 0 is a label cost function,
I T ⊆ S × L × S is the transition relation,
I s 0 ∈ S is the initial state, and
I S ? ⊆ S is the set of goal states.
We say that T has the transition hs, `, s 0 i if hs , `, s 0 i ∈ T .
A3. Transition Systems and Propositional Logic Transition Systems
Deterministic Transition Systems
Definition (Deterministic Transition System)
A transition system is called deterministic if for all states s
and all labels `, there is at most one state s 0 with s − → ` s 0 .
Example: previously shown transition system
A3. Transition Systems and Propositional Logic Transition Systems
Transition System Terminology (1)
We use common terminology from graph theory:
I s 0 successor of s if s → s 0
I s predecessor of s 0 if s → s 0
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 9 / 27
A3. Transition Systems and Propositional Logic Transition Systems
Transition System Terminology (2)
We use common terminology from graph theory:
I s 0 reachable from s if there exists a sequence of transitions s 0 − ` →
1s 1 , . . . , s n−1 − ` →
ns n s.t. s 0 = s and s n = s 0
I
Note: n = 0 possible; then s = s
0I
s
0, . . . , s
nis called (state) path from s to s
0I
`
1, . . . , `
nis called (label) path from s to s
0I
s
0−
`→
1s
1, . . . , s
n−1−
`→
ns
nis called trace from s to s
0I
length of path/trace is n
I
cost of label path/trace is P
n i=1c(`
i)
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 10 / 27
A3. Transition Systems and Propositional Logic Transition Systems
Transition System Terminology (3)
We use common terminology from graph theory:
I s 0 reachable (without reference state) means reachable from initial state s 0
I solution or goal path from s : path from s to some s 0 ∈ S ?
I
if s is omitted, s = s
0is implied
I transition system solvable if a goal path from s 0 exists
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 11 / 27
A3. Transition Systems and Propositional Logic Example: Blocks World
A3.2 Example: Blocks World
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 12 / 27
A3. Transition Systems and Propositional Logic Example: Blocks World
Running Example: Blocks World
I Throughout the course, we occasionally use the blocks world domain as an example.
I In the blocks world, a number of differently blocks are arranged on a table.
I Our job is to rearrange them according to a given goal.
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 13 / 27
A3. Transition Systems and Propositional Logic Example: Blocks World
Blocks World Rules (1)
Location on the table does not matter.
≡
Location on a block does not matter.
≡
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 14 / 27
A3. Transition Systems and Propositional Logic Example: Blocks World
Blocks World Rules (2)
At most one block may be below a block.
At most one block may be on top of a block.
A3. Transition Systems and Propositional Logic Example: Blocks World
Blocks World Transition System for Three Blocks
A3. Transition Systems and Propositional Logic Example: Blocks World
Blocks World Computational Properties
blocks states
1 1
2 3
3 13
4 73
5 501
6 4051
7 37633
8 394353 9 4596553
blocks states
10 58941091
11 824073141
12 12470162233
13 202976401213
14 3535017524403
15 65573803186921 16 1290434218669921 17 26846616451246353 18 588633468315403843
I Finding solutions is possible in linear time
in the number of blocks: move everything onto the table, then construct the goal configuration.
I Finding a shortest solution is NP-complete given a compact description of the problem.
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 17 / 27
A3. Transition Systems and Propositional Logic Example: Blocks World
The Need for Compact Descriptions
I We see from the blocks world example that transition systems are often far too large to be directly used as inputs
to planning algorithms.
I We therefore need compact descriptions of transition systems.
I For this purpose, we will use propositional logic, which allows expressing information about 2 n states as logical formulas over n state variables.
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 18 / 27
A3. Transition Systems and Propositional Logic Reminder: Propositional Logic
A3.3 Reminder: Propositional Logic
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 19 / 27
A3. Transition Systems and Propositional Logic Reminder: Propositional Logic
More on Propositional Logic
Need to Catch Up?
I This section is a reminder. We assume you are already well familiar with propositional logic.
I If this is not the case, we recommend Chapters B1 and B2 of the Theory of Computer Science course at
https://dmi.unibas.ch/de/studium/
computer-science-informatik/fs18/
main-lecture-theory-of-computer-science/
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 20 / 27
A3. Transition Systems and Propositional Logic Reminder: Propositional Logic
Syntax of Propositional Logic
Definition (Logical Formula)
Let A be a set of atomic propositions.
The logical formulas over A are constructed by finite application of the following rules:
I > and ⊥ are logical formulas (truth and falsity).
I For all a ∈ A, a is a logical formula (atom).
I If ϕ is a logical formula, then so is ¬ϕ (negation).
I If ϕ and ψ are logical formulas, then so are (ϕ ∨ ψ) (disjunction) and (ϕ ∧ ψ) (conjunction).
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 21 / 27
A3. Transition Systems and Propositional Logic Reminder: Propositional Logic
Syntactical Conventions for Propositional Logic
Abbreviations:
I (ϕ → ψ) is short for (¬ϕ ∨ ψ) (implication)
I (ϕ ↔ ψ) is short for ((ϕ → ψ) ∧ (ψ → ϕ)) (equijunction)
I parentheses omitted when not necessary:
I
(¬) binds more tightly than binary connectives
I
(∧) binds more tightly than (∨), which binds more tightly than (→), which binds more tightly than (↔)
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 22 / 27
A3. Transition Systems and Propositional Logic Reminder: Propositional Logic
Semantics of Propositional Logic
Definition (Valuation, Model)
A valuation of propositions A is a function v : A → {T, F}.
Define the notation v | = ϕ (v satisfies ϕ; v is a model of ϕ;
ϕ is true under v ) for valuations v and formulas ϕ by
I v | = >
I v 6| = ⊥
I v | = a iff v (a) = T (for all a ∈ A)
I v | = ¬ϕ iff v 6| = ϕ
I v | = (ϕ ∨ ψ) iff (v | = ϕ or v | = ψ)
A3. Transition Systems and Propositional Logic Reminder: Propositional Logic
Propositional Logic Terminology (1)
I A logical formula ϕ is satisfiable
if there is at least one valuation v such that v | = ϕ.
I Otherwise it is unsatisfiable.
I A logical formula ϕ is valid or a tautology if v | = ϕ for all valuations v .
I A logical formula ψ is a logical consequence of a logical formula ϕ, written ϕ | = ψ, if v | = ψ for all valuations v with v | = ϕ.
I Two logical formulas ϕ and ψ are logically equivalent,
written ϕ ≡ ψ, if ϕ | = ψ and ψ | = ϕ.
A3. Transition Systems and Propositional Logic Reminder: Propositional Logic
Propositional Logic Terminology (2)
I A logical formula that is a proposition a or a negated
proposition ¬a for some atomic proposition a ∈ A is a literal.
I A formula that is a disjunction of literals is a clause.
This includes unit clauses ` consisting of a single literal and the empty clause ⊥ consisting of zero literals.
I A formula that is a conjunction of literals is a monomial.
This includes unit monomials ` consisting of a single literal and the empty monomial > consisting of zero literals.
Normal forms:
I negation normal form (NNF)
I conjunctive normal form (CNF)
I disjunctive normal form (DNF)
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 25 / 27
A3. Transition Systems and Propositional Logic Summary
A3.4 Summary
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 26 / 27
A3. Transition Systems and Propositional Logic Summary
Summary
I Transition systems are (typically huge) directed graphs that encode how the state of the world can change.
I Propositional logic allows us to compactly describe complex information about large sets of valuations as logical formulas.
G. R¨oger, T. Keller (Universit¨at Basel) Planning and Optimization October 1, 2018 27 / 27